Prolactin and its receptor: H-P-C gene expression analyses
Victoria
updated 1/22/2019
Crop sac qPCR data
Some notes on this dataset: Genes-of-interest PRL and PRLR were run in duplicate for each crop sample, along with two reference genes, ACTB and rpL4. Gene expression was normalized (dCT) by subtracting the geometric mean of the reference gene duplicates (ACTB and rpL4) from the average Ct of the duplicates for each gene-of-interest.
The more positive the dCT, the lower the gene expression relative to reference genes (within a sample)
Can also think of it as “distance from reference gene”
The second normalization (ddCT) involved subtracting the average normalized expression (dCT) of the control group (here, building) from each normalized sample for that gene. More positive numbers imply lower expression than average, more negative numbers imply higher expression than average.
For this reason, it is easier to interpret the -ddCT. Now, postive numbers on -ddCT imply higher expression than average control group, and negative numbers imply lower expression than average control group, as one would expect
Lowest numbers have lowest relative expression, highest numbers have highest relative expression.
Hypothalamus
Transcriptomic data from RNAseq.
Data distribution
*Note: “Focal data” refers to data from only the five stages we are interested in: bldg, inc d3, inc d9, m.inc d8, and hatch“*
Histograms of log transformed data show
HYP PRL: possible bimodal distribution,
HYP PRLR: relatively normal distributon
Gene expression plots
GLM: Hyp PRL
ANOVA indicates a significant effect of stage, but no effect of sex or any interaction.
According to p-values in glm, inc.d3 and hatch significantly differs from manipulation. (But manip does not significantly differ from hatch).
Residual inspection: Residual plots show some outliers with large Cook’s distances and high influence. Tails of Q-Q plot are not on the line.
(data points: #64, 79, 85)
## Analysis of Deviance Table
##
## Model: gaussian, link: identity
##
## Response: log.exp
##
## Terms added sequentially (first to last)
##
##
## Df Deviance Resid. Df Resid. Dev Pr(>Chi)
## NULL 102 65.547
## stage 4 11.0749 98 54.472 0.0005504 ***
## sex 1 0.0376 97 54.434 0.7954959
## stage:sex 4 2.3786 93 52.056 0.3732991
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Call:
## glm(formula = log.exp ~ stage + sex, family = "gaussian", data = hprl)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.68291 -0.54851 0.00206 0.51655 1.90204
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.91034 0.18306 26.824 < 2e-16 ***
## stageinc.d3 -0.01854 0.23689 -0.078 0.93779
## stageinc.d9 -0.26708 0.22906 -1.166 0.24648
## stagem.inc.d8 -0.70343 0.23689 -2.969 0.00376 **
## stagehatch 0.28803 0.23689 1.216 0.22699
## sexmale 0.03822 0.14766 0.259 0.79630
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 0.5611793)
##
## Null deviance: 65.547 on 102 degrees of freedom
## Residual deviance: 54.434 on 97 degrees of freedom
## AIC: 240.61
##
## Number of Fisher Scoring iterations: 2
GLM: Hyp PRLR
“ANOVA” indicates that there is a significant effect of stage and sex on PRLR expression, but no interaction.
Manipulation day 8 is significantly higher than bldg, day 3, and day 9, but NOT significantly different from hatch.
All other stages not signficantly different from each other.
Significant sex difference, where males > females.
Residual inspection: Outliers with large Cook’s distance and influence (datapoints: #2, #87, #90)
## Analysis of Deviance Table
##
## Model: gaussian, link: identity
##
## Response: log.exp
##
## Terms added sequentially (first to last)
##
##
## Df Deviance Resid. Df Resid. Dev Pr(>Chi)
## NULL 102 13.4219
## stage 4 1.3276 98 12.0944 0.003097 **
## sex 1 3.9446 97 8.1497 5.875e-12 ***
## stage:sex 4 0.4054 93 7.7443 0.301015
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Analysis of Deviance Table
##
## Model 1: log.exp ~ stage + sex
## Model 2: log.exp ~ stage * sex
## Resid. Df Resid. Dev Df Deviance Pr(>Chi)
## 1 97 8.1497
## 2 93 7.7443 4 0.40544 0.301##
## Call:
## glm(formula = log.exp ~ stage + sex, family = "gaussian", data = hr)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.45932 -0.13168 0.03392 0.17965 0.51591
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.61608 0.07083 65.171 < 2e-16 ***
## stageinc.d3 0.16021 0.09166 1.748 0.083651 .
## stageinc.d9 0.25060 0.08863 2.827 0.005700 **
## stagem.inc.d8 0.34172 0.09166 3.728 0.000325 ***
## stagehatch 0.25171 0.09166 2.746 0.007188 **
## sexmale 0.39148 0.05713 6.852 6.72e-10 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 0.08401782)
##
## Null deviance: 13.4219 on 102 degrees of freedom
## Residual deviance: 8.1497 on 97 degrees of freedom
## AIC: 45.017
##
## Number of Fisher Scoring iterations: 2
Pituitary
Transcriptomic data from RNAseq.
Data distribution
Histograms of log transformed data show
PIT PRL: possible bimodal distribution, w/ one large outlier,
PIT PRLR: relatively normal distributon
##Gene exp. plots 
GLM: Pit PRL
“ANOVA” : significant effects of stage and sex, with no interaction.
Manipulation day 8 is significantly higher than bldg, day 3, and day 9, but NOT significantly different from hatch.
All other stages not signficantly different from each other.
Significant sex difference, where males < females.
Residual inspection: Outliers with large Cook’s distance and influence (datapoints: #8, #28, #78)
## Analysis of Deviance Table
##
## Model: gaussian, link: identity
##
## Response: log.exp
##
## Terms added sequentially (first to last)
##
##
## Df Deviance Resid. Df Resid. Dev Pr(>Chi)
## NULL 103 156.292
## stage 4 86.123 99 70.169 < 2.2e-16 ***
## sex 1 4.733 98 65.437 0.006628 **
## stage:sex 4 5.085 94 60.351 0.094528 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Analysis of Deviance Table
##
## Model 1: log.exp ~ stage * sex
## Model 2: log.exp ~ stage + sex
## Resid. Df Resid. Dev Df Deviance Pr(>Chi)
## 1 94 60.351
## 2 98 65.437 -4 -5.0853 0.09453 .
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
GLM: Pit PRLR
“ANOVA” : significant effects of sex only Manipulation day 8 is significantly different from hatch and building, but not incubation stages.
Significant sex difference, where males > females.
Residual inspection: Outliers with large Cook’s distance and influence (datapoints: #5, #18, #45)
## Analysis of Deviance Table
##
## Model: gaussian, link: identity
##
## Response: log.exp
##
## Terms added sequentially (first to last)
##
##
## Df Deviance Resid. Df Resid. Dev Pr(>Chi)
## NULL 103 22.624
## stage 4 1.0198 99 21.604 0.2111
## sex 1 4.7463 98 16.858 1.836e-07 ***
## stage:sex 4 0.4545 94 16.403 0.6260
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1## Analysis of Deviance Table
##
## Model 1: log.exp ~ stage + sex
## Model 2: log.exp ~ stage * sex
## Resid. Df Resid. Dev Df Deviance Pr(>Chi)
## 1 98 16.858
## 2 94 16.403 4 0.4545 0.626
Crop sac
Data distribution
Histograms of log transformed data show
Crop PRL: relatively normal, with ~ 4 low outliers
Crop PRLR: relatively normal distributon
Crop sac gene expression plots
GLM: Crop PRL
“ANOVA” : very weak trend towards an interaction, but not really anything else there. Not really anything going on in downstream analyses… Discuss mostly the fact that PRL gene expression was detectable in the crop.
Residual inspection: Outliers with large Cook’s distance and influence (datapoints: #29, #35, #85)
##
## Call:
## glm(formula = log.exp.neg ~ Stage + Sex, family = "gaussian",
## data = cp)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -1.9239 -0.2818 0.1484 0.3413 0.9717
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.88335 0.19697 9.562 2.71e-13 ***
## StageInc_D3 -0.11143 0.25184 -0.442 0.660
## StageInc_D9 -0.04919 0.25276 -0.195 0.846
## StageManip_D8 0.04144 0.31990 0.130 0.897
## StageHatch -0.05025 0.24941 -0.201 0.841
## Sexm 0.23857 0.16535 1.443 0.155
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 0.4085906)
##
## Null deviance: 23.443 on 60 degrees of freedom
## Residual deviance: 22.472 on 55 degrees of freedom
## (34 observations deleted due to missingness)
## AIC: 126.2
##
## Number of Fisher Scoring iterations: 2
PRLR Crop
“ANOVA” : weak trend towards stage Not really anything going on in downstream comparisons. Trend towards hatch being significant (may be driving the weak trend towards an effect of stage)?
Residual inspection: Outliers with large Cook’s distance and influence (datapoints: #17, #62, #83), some bad outliers on this one!
## Analysis of Deviance Table
##
## Model: gaussian, link: identity
##
## Response: log.exp.neg
##
## Terms added sequentially (first to last)
##
##
## Df Deviance Resid. Df Resid. Dev Pr(>Chi)
## NULL 87 27.303
## Stage 4 2.4873 83 24.816 0.08285 .
## Sex 1 0.0927 82 24.723 0.57924
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1##
## Call:
## glm(formula = log.exp.neg ~ Stage + Sex, family = "gaussian",
## data = cr)
##
## Deviance Residuals:
## Min 1Q Median 3Q Max
## -3.3939 -0.0512 0.1041 0.2560 0.9704
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.91299 0.14432 13.255 <2e-16 ***
## StageInc_D3 0.12526 0.18117 0.691 0.4913
## StageInc_D9 0.05969 0.18949 0.315 0.7535
## StageManip_D8 0.03419 0.19451 0.176 0.8609
## StageHatch 0.45468 0.18343 2.479 0.0152 *
## Sexm -0.06553 0.11818 -0.554 0.5807
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## (Dispersion parameter for gaussian family taken to be 0.3015003)
##
## Null deviance: 27.303 on 87 degrees of freedom
## Residual deviance: 24.723 on 82 degrees of freedom
## (7 observations deleted due to missingness)
## AIC: 152.01
##
## Number of Fisher Scoring iterations: 2
